Deleting the ten-minute polling loop left the "auto upload" switch in settings with no
consumer - the operator could turn it on and nothing would ever act on it, which is worse
than the loop it replaced.
Uploading now happens when a contact is saved. That is what the switch always meant, and
doing it at that moment means somebody is present to see the outcome: a partial failure
says so, and the QSOs that did not go stay in the queue for a manual upload from settings.
The loop reported nothing at all - a grid mismatch hit an empty if block and every other
failure retried forever in the background.
The upload goes through the view model rather than the composable, so the log screen still
touches no repository.
WaveLogApi decided an upload had succeeded from the HTTP status alone. Wavelog validates
after responding, so a rejected QSO comes back as 200 with `{"status":"failed","reason":
"..."}` - and the uploader then called markUploaded and dropped it from the queue. The
contact was lost and the operator was told the upload succeeded.
Response shapes are transcribed from the Wavelog API reference, not guessed: success is
`status: success` or `successful`, a duplicate is `status: dupe` with a 200, failures are
`status: failed` with `reason` or `status: error` with `message`.
WavelogResponse reads the body. Four outcomes: accepted and duplicate both clear the
queue entry, because the log holds the QSO either way; rejected keeps it and surfaces the
server's own explanation; and a status field we cannot recognise also keeps it, since
costing a retry beats losing a contact. Parsed as text rather than with JSONObject because
org.json is compileOnly in core:domain and a JVM test would otherwise assert against a
stub. Whitespace around separators is collapsed before matching - a first attempt listed
spacings and missed `{ "status" : "failed" }`, which a probe caught.
Two other things in the same area.
The ten-minute auto-upload loop is gone. It retried the queue in the background with no
way to tell the operator anything: a grid mismatch was swallowed by an empty if block and
every other failure retried silently forever. A QSO that cannot be uploaded now waits for
a manual upload from settings, where the result is actually shown.
The upload path no longer builds user-facing text in Kotlin. UploadOutcome carried a
pre-formatted Chinese string, so the message ignored the device language whatever the
locale files said. It now reports a Reason the view model maps to resources, which needed
a format-argument overload on IShowToast to get a count into a localised message.
The previous commit listed the refusal wordings it knew - "invalid login" and "login
denied" - and skipped everything else as chatter. That list was incomplete. Probing the
parser against responses captured from live servers found three it missed:
# Login by user not allowed observed on rotate.aprs2.net
# Port full
# Server full
Each was skipped as a keepalive, so the login timed out into Unknown, Unknown is
deliberately read as "may be working", and every send afterwards reported success to an
operator the server had refused. Exactly the failure the previous commit fixed, reached
by a different wording.
Inverted: identification and keepalive comments are recognised positively, and anything
else the server says during login counts as an objection. The trade is that an unforeseen
harmless comment would read as a refusal - but that errs towards reporting failure rather
than claiming success, which is the direction this feature has been wrong in throughout.
The keepalive prefixes come from a live capture rather than guesswork. aprsc repeats its
own identification with a timestamp every twenty seconds:
# aprsc 2.1.21-gbfc2090 25 Aug 2026 16:41:07 GMT T2UK 195.201.15.71:14580
Two tests had invented a `# Tue Aug 25 ...` date line and a `# keepalive N`, neither of
which any server sends. Both now use the captured format.
Also here: the QRZ cookie test in settings goes through the repository instead of
scraping from the UI. It was the last caller of QrzGridClient, which is deleted, and it
built its result from hardcoded Chinese strings inside the composable - those move to
resources, and the four outcomes are now distinguished, where before an expired cookie
and a station with no grid on file produced the same message.
LogTab read the QRZ cookie straight out of SharedPreferences through LocalContext,
inside composition, on every submission - disk access in a composable, around the
repository layer, with the client referenced by fully-qualified name inline. And it
did `if (grid != null)`, so a lookup that failed for any reason left the QSO without
a grid and told the operator nothing.
IQrzGridLookup lives in core:domain, QrzGridLookup in core:data owns the cookie read,
and the view model exposes lookupGrid. The composable now takes a callback and handles
each outcome: a locator is attached, no locator on file passes quietly because nothing
is wrong, an expired cookie says to paste a fresh one, and an unreachable QRZ says so.
That is what the four-outcome QrzGrid type from ce68f487 was for - until now nothing
consumed it and the old nullable client was still the one being called.
Note for anyone extending RadarScreen: the local holding the view model cannot be
referenced as `viewModel` inside a lambda, because that name also resolves to the
composable factory function. Hence the explicitly typed local.
The old QrzGridClient is now unused here but left in place; removing it belongs with
the settings screen, which still calls it to validate a pasted cookie.
`submit()` opened with `if (call.length < 3) return`. During a pass the operator typed
a callsign, pressed done, and nothing happened - no entry, no message, no way to tell
the app had decided against them. None of the logging software surveyed for this work
- N1MM+, DXLog, PoLo, HAMRS - discards a submission silently.
Validation is deliberately loose, because strictness costs more than it saves. Checked
against 28 real callsigns, a typical strict pattern rejects 16 of them: W1AW/4, 2E0ABC,
9A1CCY and SV2ASP/A among others. A pattern permissive enough to accept those also
accepts a Maidenhead locator as a callsign. There is no regex that catches typos without
throwing away legitimate calls, so CallsignEntry rejects only what cannot be a callsign
- empty, one character, illegal characters, all digits, all letters - and reports doubt
as a warning that still logs the contact.
Two warnings exist. A six-character grid-shaped entry says so, because grid and callsign
are exchanged together on FM satellites and the fields sit side by side. A station already
worked this pass says so too, without blocking: the same station on a later pass is a
legitimate new contact, and contest loggers default to working duplicates - DXLog
describes refusing them as an outdated habit.
That warning also replaces the duplicate suppression, which was a 300ms window comparing
the last callsign, admitted in its own comment to be a workaround. It could silently
discard a real second contact, and a set of calls worked this pass is both honest and
more useful. It survives configuration changes via rememberSaveable.
Not addressed here: the QRZ grid backfill still reads the cookie out of SharedPreferences
from inside a composable through LocalContext, and still reports nothing when a lookup
fails. IQrzGridLookup is added for that, but wiring it needs the container, the view model
and the UI to change together.
The settings card had a "Compute passcode" button that derived the value from the
callsign and filled the field in. It was added by request, so it stayed while the
previous commit removed the same derivation from the connection path - which left
the app contradicting itself: the background no longer invented a passcode, but the
UI still offered to.
APRS-IS treats the passcode as a licence check and states that supplying it to a
user is the software author's responsibility. APRSdroid carries the identical
algorithm in the same source file and deliberately does not use it for this, opting
to validate what the operator typed and link out to request one. Filling the field
in claims a check that nobody performed.
The button now opens the passcode request page. AprsPacket.passcode stays in
core:domain because validating an entry means recomputing the expected value, and
its import is dropped from the card, which no longer needs it.
LoTW refuses a QSO whose SAT_NAME is not spelled as its accepted list has it - its
own help page gives AO7 against AO-7 as a rejection - so the name we upload decides
whether a contact can ever be confirmed. The old code derived it with
substringBefore('('), which returns the descriptive half of a TLE name rather than
the OSCAR designator: measured against live Celestrak amateur data, 0 of 96
satellites resolved to something LoTW accepts. ASRTU-1 went up as ASRTU-1 where
LoTW wants AO-123.
Keying on the name cannot be made to work, because the sources disagree. Of the 49
satellites carried by both Celestrak amateur and AMSAT nasabare, 33 are named
differently - 43017 is RADFXSAT (FOX-1B) in one and AO-91 in the other, 43700 is
ES'HAIL 2 against QO-100 - so which name a QSO got depended on where the operator
fetched their TLE. The catalogue number is identical everywhere, so the table is
keyed on it and OrbitalPass.catNum is now threaded through to the QSO and persisted.
The name path stays as a fallback for contacts logged before the number was
recorded, and got two fixes of its own: it tries either side of the parentheses
rather than assuming the designator is on the left, and tolerates a differing
separator so RADIO ROSTO (RS15) reaches RS-15. Resolution now returns the list's own
spelling, so Arsene is not uploaded as ARSENE and rejected the same way AO7 would be.
Measured on the same data: 3 names resolved before, 20 by name alone now, 30 with
the catalogue number, and no satellite that used to resolve stopped resolving.
The table gained the nine TEVEL-2 satellites after an audit found them missing.
Every source writes those TEVEL2-N while LoTW has TEV2-N, which stripping separators
does not bridge - TEVEL21 is not TEV21 - so they resolved to nothing at all. They
launched in 2025 and are workable now. Their numbering is not sequential: 63217 is
TEVEL2-1 while 63213 is TEVEL2-4.
All 38 entries were cross-checked two independent ways: every catalogue number
appears in the app's own configured sources under a name consistent with the LoTW
spelling, and ARRL's startDate for each name agrees with the launch year in the
TLE international designator - which is what would catch a number pointing at the
wrong object, since a name can match by luck. Nothing here was typed from memory;
an early hand-written draft had AO-123 as 62690 when it is 61781.
Two corrections to 10c415fa and 0889a3bd, keeping what those got right and
undoing what they cost.
The transcript now follows new text through an explicit follow flag rather than
comparing scroll position against maxValue. maxValue is written during layout,
after the composition that would read it, so the comparison tested the previous
frame's height: following fell progressively short of the true bottom and, once
the gap passed the slack, latched the operator out of follow-mode until they hit
the exact end. Scrolling away still stops it, which is the point.
The AMSAT day cell goes back to 28 dp. Raising it to 48 dp for the minimum touch
target measured a 71% increase in row pitch - 14 satellites per screen down to 8
on a 6.1" phone - and comparing many satellites at a glance is what that page is
for. Compose cannot extend a touch target past the layout bounds, so this is a
choice rather than a fix; 28 dp is also what shipped before, so the regression
was mine. The contentDescription added alongside it stays, since it costs nothing.
The mixer runs above unity for any ordinary input - the Hilbert kernel's L1 gain
is 2.51, so amplitude 0.7 peaks at about 1.76 - and the output was hard clipped
to fit. Clipping squares the waveform off and generates odd harmonics, which the
widened waterfall would now put on screen.
Measured, the harmonics happen to be harmless today: TARGET_HZ is a quarter of
the sample rate, so 3f, 5f, 7f and 9f all fold back onto the tone itself and
out-of-band energy stayed at 0.00%. That is a coincidence between two constants,
not a property of the design. At a 700 Hz target the third harmonic folds to
1100 Hz - inside the analysis window, where no filter may remove it and the model
would read it as a second tone.
So two changes, because neither alone is enough. A peak-following gain scales the
mixer output to fit rather than clipping it: measured 0 of 3200 samples on the
rail, against a clipped waveform parking there for much of every cycle. And a
95-tap windowed-sinc band-pass over the model's window removes whatever the mix
leaves outside it - images, harmonics, the far sideband - measured at 58-60 dB
rejection with 0.09 dB of passband ripple and out-of-band energy down to 0.0002%.
The gain is shared across chunks so it cannot step at a boundary, and the filter
carries tap history for the same reason the Hilbert filter already did.
The band-pass adds 47 samples of linear-phase group delay, 14.7 ms, which delays
the keying envelope without distorting it - 4% of a dot at 40 WPM.
CwToneShifterStreamingTest's boundary criterion was wrong, and the band-pass
exposed it: distanceToBoundary measured only forward, so the first samples of a
chunk came out 320 away from "the" boundary and counted as interior when they are
the far side of the same seam. Both filters need samples ahead of the output they
are producing - 32 for the Hilbert transform, 47 for the band-pass - and with the
distance measured to the nearest boundary either way, interior divergence is
0.000116 against a 0.01 budget.
Also: the CW transcript now follows the newest text, but only while the operator
is already at the bottom, so scrolling back to read earlier traffic is not undone
by the next decoded character.
The waterfall showed only the model's 400-1200 Hz window, so a tone outside it
was absent from the picture entirely. Measured on keyed audio, the brightest
column in that narrow view swings 1.01x between key-down and key-up against
13.76x for a tone in range - it carries no keying at all, so the operator could
not tell a signal was present, let alone where it was. Markers alone could not
fix that: they pointed at a frequency with nothing drawn there.
compute() now takes an optional bin range, defaulting to the model's own, so the
decoder path is byte-identical and the golden-vector test still holds. The
display asks for DC to Nyquist, 129 bins against 65. The FFT already computed
every bin - this only changes which are kept - so the cost is a wider copy.
The decoder window is framed and faintly lifted, since half the picture is now
outside what the model reads and nothing said which half.
Marker fixes found while reviewing the render: the tone marker was orange, which
is a colour the inferno ramp itself passes through, so a marker sitting on the
trace it pointed at was indistinguishable from the keying gaps in that trace -
invisible in exactly the case it existed for. It is cyan now, and both markers
are pips in a gutter above the spectrum rather than lines across it.
Also from the release audit:
- compute()'s bin-count guard was written as a three-term disjunction, which any
custom range satisfies regardless of bin count, leaving the model invariant
unenforced for the caller most able to break it. Rewritten as an implication,
with a Nyquist bound so no range can index past the FFT output.
- signalStrength was gated on a confirmed out-of-window tone, which is false when
detection fails - and it fails for a slow fist, measured at prominence 2.5
against a 4.5 threshold for 15% duty. So the meter still read half scale beside
an empty transcript. It now requires a tone confirmed decodable: 11 flow
combinations, 3 wrong before, 0 wrong after.
- detectedToneHz never expired, so after retuning into the band the hint kept
naming the frequency the operator had left, indefinitely. It now clears after
10 s without a tone, which is clear of any real gap - the longest being 1.7 s
between words at 5 WPM.
- The waterfall label read estimatedPitch while the hint read detectedToneHz, two
numbers up to 800 Hz apart both claiming to be the tone. Both read the latter.
- Removed a redundant toFloat() that the compiler warned about.
Accessibility, untouched until now: the waterfall was a bare Canvas and the AMSAT
day cells bare Boxes, so both announced nothing at all - on the status page that
is the entire content of the screen. Both now carry a contentDescription naming
the tone or the day's worst status and report count. The AMSAT tap target goes
from 28 dp to 48 dp with the coloured tile still 28 dp, so the grid keeps its
density. Strings in all nine locales for both modules.
Opinion split on the stripes, so Settings > Other now has a switch. On by
default, since the flat tile it replaced hid intra-day outages, which is the
problem the stripes were introduced to solve.
Flat mode is deliberately not the old behaviour. The old cell took its colour
from the first slot with a report and its count from that same slot, so a day
that worked in the morning and failed all afternoon read as "worked" - measured
across eight representative day shapes, two of them had their failure hidden
outright, and the count reported 1 where the day held 24 reports. Flat mode now
takes the day's worst status and the day's total count, so the summary can
understate detail but not hide bad news. The help text says so, in case someone
turns the switch off expecting the tile they remember.
The count is drawn in black or white by relative luminance rather than always
white: on the telemetry amber, white measured 1.83:1 against WCAG's 3:1 for
large text, and that cell does carry a count whenever a day held nothing but
telemetry reports. All six status colours now clear 3:1, the worst being 3.03.
SatStatusViewModel collects the setting rather than reading it once - the switch
is on another screen, so the operator is always elsewhere when they change it
and would otherwise return to the old style.
Strings in all nine locales.
With tone shift off and the operator tuned outside 400-1200 Hz, the page did not
go quiet - it went confidently wrong. Three measurements, all reproduced against
the real spectrogram path:
estimatedPitch is (32 + loudestBin) * 12.5 - shiftHz with the bin confined to
0..64, so with no shift applied it can only ever report 400-1200 Hz. It cannot
express 1500 Hz, and it does not try: it publishes whichever window edge the
leakage piles against. For a 1500 Hz tone that is 1200 Hz.
That leakage is not faint. The waterfall normalises to the loudest value on
screen, so 50 of 65 bins clear the 0.06 draw threshold and the picture shows a
keyed-looking column pinned to the right edge - the 1200 Hz column runs 25 times
the 400 Hz one.
signalStrength is prominence over the window mean, so the same leakage scores
0.78 and paints the meter to 78% of full width.
So the operator got a strong-signal bar, a plausible 1200 Hz readout, a picture
that looked like a signal, and an empty transcript, with nothing saying why.
The scan that can see past the window now runs whether or not shifting is
enabled - it is the only measurement that can - and publishes through a new
detectedToneHz flow kept separate from estimatedPitch. Overloading the latter is
what let the 1200 Hz claim out in the first place, so the two meanings stay in
two flows. The shift decision still only happens when the setting is on. Cost is
one 121-bin scan every 2 s.
The meter now reads zero when a tone is out of range and not being shifted in: it
is a claim that something decodable is present, and in that state nothing is.
A line under the waterfall says which case the operator is in - the tone was
moved in, or it is out of range and tone shift is off, naming the frequency and
the remedy. Strings in all nine locales; feature:cw only had five, so values-es,
values-ru, values-si and values-uk are new, with the Turkish apostrophe escaped.
CwToneShifterTest pins the premise the hint rests on: that the scan reports tones
the model window excludes, at 120, 250, 1400 and 1500 Hz.
The guard suppressed every marker, the target line included, whenever the
reported pitch was not positive. Shifting a low tone UP makes that routine:
pitch is (loudestBin * 12.5 - shiftHz), so with a 100 Hz tone shifted +700 Hz it
goes negative for 25 of the 65 bins, down to -300 Hz, and updateSignalMetrics
applies no prominence test so mains hum in a key-up gap is enough to park the
argmax down there. 77 reachable (tone, bin) pairs across 100-350 Hz produce it.
The result was the display showing nothing at all while the shift was active -
exactly what the previous commit set out to fix.
The target line is now drawn on the strength of the shift alone, since a shift
being applied is the fact worth showing and it does not depend on the pitch. A
non-positive pitch marks the low edge, which is where such a tone actually is,
and only the numeric label is suppressed because the number itself is nonsense.
A NaN pitch previously slipped past all three comparisons and rendered the HIGH
edge marker labelled "0 Hz"; it now draws the target line only.
TONE_SHIFT_TARGET_HZ reads CwToneShifter.TARGET_HZ instead of recomputing the
window midpoint. The two are equal today by coincidence, not construction:
retuning either would leave the green line marking a frequency nothing is
delivered to, silently. CwToneShifterTest now pins TARGET_HZ inside the window
and clear of its edges, which is the one part of this the JVM suite can hold.
The label side now tips at the target rather than the window maximum, so a pitch
sitting on the upper edge gets its text on the same side as its line.
The edge marker was drawn outward from the canvas edge, so every one of its
three line segments fell outside the clip and nothing rendered. Measured at a
typical 320 px width: 0 of 3 segments visible on either side. That is the one
case the marker exists for - an out-of-window tone is absent from this picture
by definition, so with the marker clipped away the operator has no signal at all
that a shift is happening. Which is what was reported.
It is now a solid bar along the edge the tone lies beyond, plus a chevron whose
arms open inward from it, so the whole marker sits inside the clip while still
reading as pointing off-picture.
Three further defects in the same code:
The frequency label was pinned to TopStart while its background rect tracked the
tone's frequency, so at 1500 Hz the rect sat at x=278 and the text at x=11. The
rect is gone and the label now sits on whichever side the marker is on.
Markers were drawn after two early returns that fire on an empty or silent
spectrum. A shift is deliberately held through key-up gaps, so the markers were
blinking out during the very silences the shift survives. They now draw
unconditionally, after the spectrum so it cannot bury them.
dashCount floored, leaving up to 8 px of the column undrawn at the bottom.
Also extracts the marker drawing into a DrawScope extension, hoists the shared
colours and the target frequency to file-level constants, and rounds the label
instead of truncating it.
The Canvas marker was fixed to draw at estimatedPitch, but the overlay Text
still computed its label from estimatedPitch + toneShiftHz, which showed the
shifted position (800 Hz) instead of the original tone (e.g. 1500 Hz).
estimatedPitch is already corrected back to the original tone frequency
(the spectrogram computes from shifted audio, and updateSignalMetrics undoes
the shift), so adding toneShiftHz to it again placed the orange marker at the
shifted position - right on top of the green target line, making them
indistinguishable.
The orange marker now goes directly on estimatedPitch. When the original pitch
is outside the visible 400-1200 Hz band, an arrow at the nearest edge points
toward it instead.
When the tone-shift feature moves a tone into the model's 400-1200 Hz window,
the waterfall now shows two visual markers so the operator can see what is
happening: a green dashed line at the target (800 Hz) and an orange frequency
label at the top-left showing the original pitch.
The waterfall draws the RAW audio, not the shifted audio, so a 1500 Hz tone was
always invisible regardless of the shift setting. The markers close the gap:
the operator can now see that a tone was detected and where it was moved, even
when the original pitch is outside the visible band.
activeShiftHz is now a StateFlow exposed through ICwDecoder so the UI can
observe it without polling.
The API caps at 500 records regardless of the hours requested. With 88 catalog
satellites, eight of them more active than 50 reports per 72 hours, quieter
satellites get crowded out. Measured live: the global pull returned 500 reports
covering 36 satellites, while the summary endpoint reported 743 reports across 38
satellites. 26 of 38 satellites had incomplete data, and two (PO-101_[FM] and
TEVEL2-6_[FM]) had zero reports in the global pull despite having reports in the
summary.
The summary endpoint (api/v1/summary.php) returns per-satellite report counts in
one request, so the fix adds one extra call rather than the 88-request
alternative of per-satellite pulls. A satellite whose global pull is incomplete
gets a subdued "68 / 116" marker next to its name, telling the operator the page
knows there is more data it could not fetch. The marker is silent when the
summary is unavailable or the counts match, so the feature degrades gracefully.
The earlier no-data grey (0xFFE8E8E8) already prevented the worst case: slots
crowded out of the global pull were marked as "we never looked" rather than
claiming "nobody reported". The marker now closes the remaining gap: the page
can honestly say "we know there are 116 reports for this satellite but we could
only show you 68 of them".
Also fixed a subagent mutation-testing residue: the coverage floor had been
moved from global (reports.minOfOrNull) to per-satellite (satReports.minOfOrNull)
and left in the tree. One test caught it (coverage is judged from all reports,
not one satellite's), proving the test has teeth.
Adds getAmSatSummary to IRemoteSource and RemoteSource, parseSummary to
AmSatRepository, and summaryCount to SatStatus. All eight test-file
implementations of IRemoteSource were updated for the new method.
Grey meant two different things. The API caps at 500 records however many hours
are requested: measured against the live endpoint, a 72-hour request returned 500
reports spanning only 49 hours, so the oldest 9.5 hours of the third day had no
data at all. Those cells were painted the same grey as "nobody reported", which
claimed knowledge we did not have - 352 of 3168 cells on a real page, a third of
the third day's column.
Slots entirely older than the earliest report in the response now use a lighter
grey. Coverage is judged from all reports rather than per satellite: a quiet
satellite has no reports of its own, but the slots it shares with the rest of the
response were still covered, so it must read as "not heard" rather than "unknown".
The two greys are now in the legend, which previously listed only the four active
states. That matters more than it sounds: on the live page 81% of cells are
"nobody reported" and 11% are outside our data, so a user looking at a mostly-grey
row had no way to tell a dead satellite from a gap in what we fetched. The legend
chips use a solid dot, so the two greys stay distinguishable despite the 25%
alpha background. Strings added to all nine locales.
Three tests cover it: a day entirely before the data starts, a day straddling the
boundary, and an empty response marking nothing as covered.
Two defects in our own AMSAT page, both found by auditing the change that exposed
them.
The day columns claimed to be dates but were a rolling window anchored on the
fetch time. Fetching at 06:07 UTC put 17.9 hours of yesterday into the cell
labelled today; measured against a live amsat.org page of 1021 reports, 73% of
them landed in the wrong day column and none matched the official cell. Days are
now UTC calendar days and slots are fixed UTC bands - slot 0 is 22:00-24:00, slot
11 is 00:00-02:00 - so a cell's contents match its label whenever it is fetched.
The day cell painted one colour for the whole day, taken from the first slot that
had a report, so a satellite that worked all morning and failed all afternoon
looked identical to one that worked once - the reported symptom. It now draws one
stripe per two-hour slot in the same 64x28 dp footprint. Twelve stripes are about
5 dp each, roughly 15 px at 440 dpi, and runs of the same status merge visually,
so a day reads as a few blocks rather than twelve lines. Every density from ldpi
up allocates all twelve without dropping one, and the 4 dp corner radius leaves
95% of the end stripes visible. The report count text is gone; tapping a day
still lists every report from it, which was already the richer view.
buildStatuses and ApiReport are internal rather than private so the grid contract
can be tested. AmSatSlotBuildTest drives it directly: fetchStatus cannot be
tested here because the parsing around it uses Android's JSONObject, a JVM stub
that makes every call return null - eight of nine tests written against it failed
for that reason before being rewritten.
Also corrects three KDoc comments claiming 5 days when the code builds 3, and
records in AGENTS.md that the status colours are ARGB literals in core:data,
duplicated in MainTheme, which anything needing themeable or colour-blind-safe
colours has to fix first.
DeepCW only analyses 400-1200 Hz - its input tensor is 65 bins wide, fixed at
training time - so a CW note outside that range is invisible to the decoder.
This adds an opt-in preprocessing step that moves such a tone to 800 Hz, the
window centre, extending the usable pitch range without touching the model.
Single-sideband mixing via a 63-tap Hilbert transformer. Plain real mixing was
measured and rejected: shifting 1500 Hz to 800 Hz left a fold-back image at
1000 Hz at 0.999 of the wanted amplitude, inside the window. Zero-stuff
upsampling plus lowpass handled downward shifts but left a 0.996 image when
shifting 300 Hz upward. The Hilbert approach measures clean on nine tones from
150 to 1550 Hz: one peak at the target, nothing above 0.3 relative amplitude.
In-window energy for a 1500 Hz input goes from 6.8% to 94.6%.
Only out-of-range audio is processed. A tone already inside 400-1200 Hz is
returned untouched (same array instance, no copy), and with the setting off the
audio path is exactly what it was before.
CwToneShifter.Streaming carries the Hilbert filter history and mixer phase
across capture chunks. Shifting each chunk in isolation left 62 of every 320
samples convolving against zeros, inflating envelope ripple to 8.7x the
whole-buffer baseline. A residual difference in the last ~3 samples of each
chunk is causal and documented: those output samples would need input that has
not been captured yet.
Detection pools chunks rather than gating on one. A capture chunk is 4410
samples at 44.1 kHz but only 320 after resampling to 3200 Hz, so requiring
1280 samples in a single chunk would have made the feature dead code - the two
independent audits both found this before it shipped. Detection now runs on a
pooled 0.4 s window, at most every 2 s.
Toggling the setting or a change in the detected shift drops the buffered
audio: the 20 s window would otherwise keep decoding samples moved by the old
amount, and the pitch readout could only be correct for one of them. The
readout itself subtracts the active shift so it shows the pitch on the radio,
not the shifted one.
Settings: OtherSettings.cwToneShiftEnabled, off by default, persisted and read
back in SettingsRepo, toggled from the Other card in Settings with a help line
explaining the 400-1200 Hz limit. Strings added to all nine locales. The
decoder reads the flag per chunk, so the toggle applies without restarting
capture.
Debug: the enabled-state transition, each detection verdict (no tone / inside
window / shifting by N Hz), and every shift change are logged, with the noisy
paths throttled to the 2 s detection interval. CwProbe records shift changes
only, keeping well inside its 1 MiB cap.
Tests: 8 shifter tests (detection sweep, noise rejection, pass-through
identity, image-free shifting across 8 tones, end-to-end spectrogram energy),
8 streaming tests (chunk continuity, history retention, reset semantics, chunk
sizes above and below the history window), and 6 gate tests including a
regression guard that a 320-sample chunk must be able to reach the detection
threshold. All 53 CW tests pass, golden vectors included.
Post-merge audit found code the merge left unreferenced:
- SettingsRepo: keySatelliteUrls / keyTransceiversUrls / separatorUrl were
upstream's list-shaped data-source keys. The merge kept the fork's map-shaped
DataSourcesSettings, so these three had a definition and zero uses.
- feature/status/res/drawable/ic_refresh.xml: SatStatusScreen imports
core.presentation.R only, so its R.drawable.ic_refresh resolves to the
core copy; the feature-local copy was never addressable. It was the only
file under feature/status/src/main/res, so the directory goes with it.
Verified zero references with a repo-wide grep before removing each symbol.
compileReleaseKotlin plus core:domain / core:data / feature:map /
feature:roaming unit tests stay green.
findMutualPassesFallback skips a candidate with `if (refinedLos <= refinedAos)
continue` but left searchStart untouched, so the next loop iteration called
findNextMutualPass with the same start time and received the same pass again.
Today that branch is theoretically unreachable: refineEdge's 70 s window always
spans findNextMutualPass's 60 s sampling step, so the refined AOS/LOS can only
move inward and never cross. But the invariant is fragile - any future change
to the search step or the refine window (or a near-horizon pass whose
crossings land at the window edges) makes the loop spin forever on one pass,
freezing the query coroutine.
Advance searchStart past the collapsed pass before skipping, breaking the
cycle regardless of how the windows shift. Behaviour for the current reachable
paths is unchanged.
RoamingScreen carried ~170 lines of decompiled range-lookup tables
(encodeLon/encodeLat) that re-implemented exactly what core:domain's
positionToQth already does. A probe calling both over 16,471 sampled
coordinates (every 2 degrees across the full globe) found byte-identical
8-char locators, so the tables were pure duplication - two implementations of
the same Maidenhead encoding that had to be kept in sync (the earlier boundary
fix had to be applied twice).
Replace them with positionToQth and split its standard-ordered output
(lonField latField lonSquare latSquare lonSub latSub lonSubsub latSubsub) back
into the per-axis segments the UI consumes: the 3x3 ring (first 4 chars),
markerLeft (lon subsquare), markerTop (lat subsquare). Out-of-range input
keeps the old blank-segment behaviour: positionToQth returns null, the locator
becomes spaces, qthNeighbors returns an empty ring, and the marker lookups
fall back to 0.
Net -160 lines. All existing RoamingState tests and the new equivalence probe
pass; :feature:roaming compiles.
RemoteSource's five suspend functions and SatStatusViewModel's two fetch paths
caught bare Exception, which also swallows CancellationException. When the
owning scope is cancelled (screen leaves, app closes) a cancelled network call
was reported as a null/error result instead of stopping: the caller kept
running until the next suspension point, and SatStatusViewModel wrote state
updates into an already-cancelled scope. Correct coroutine hygiene is to let
cancellation propagate - rethrow CancellationException before the generic
catch. Verified semantically with an asyncio probe: a swallowed cancel returns
a normal-looking null and the caller continues; a propagated cancel stops the
coroutine immediately.
No behaviour change for real errors; :core:data and :feature:status compile.
The map info panel chose which pass to describe with
allPasses.find { it.catNum == catnum && it.progress < 1 }
but OrbitalPass.progress defaults to 0 and nothing in the repository or the
prediction layer ever assigns it - PassesViewModel computes progress on its own
local copy of the list and never writes it back. The predicate was therefore
always true, so the lookup returned the satellite's *first* pass forever.
Simulated over an ISS timeline with three passes (10:00, 12:00, 14:00), 4 of 6
sampled instants were wrong: from 10:30 onwards the panel still pointed at the
10:00 pass with its countdown frozen at 00:00:00, instead of counting down to the
12:00 and 14:00 passes. Only re-fetching the pass list refreshed it.
Select by time now: the pass currently in progress, otherwise the earliest one
still upcoming. Extracted as the pure internal selectCurrentOrNextPass so it is
testable, with the reasoning recorded so the progress field is not reintroduced
as a filter here.
Restoring the old predicate fails 5 of the 6 new tests; with the fix
:feature:map:testDebugUnitTest, :core:domain:test, :core:data:testDebugUnitTest
and :feature:roaming:testDebugUnitTest are all green.
The ground track is cut into polylines so none of them spans 180 degrees, but the
cut only ever appended the outgoing edge point. The next polyline therefore began
at the first sample past the meridian - typically around -178 - so the drawn
track stopped at the edge on one side and reappeared inland on the other,
leaving a visible gap on every orbit that crosses the Pacific.
The edge point also reused the *next* sample's latitude, so the closing leg
jumped: for 179 -> -178 spanning 14 -> 16 degrees latitude the edge was placed at
16.0 instead of the true crossing at 14.667.
Now a crossing closes the current polyline on the edge it leaves through and
opens the next one on the opposite edge, both at the interpolated crossing
latitude, so the seam is continuous.
The split is extracted as the pure internal splitAtAntimeridian/crossingLatitude
pair, which also gives feature:map its first unit tests. Verified against a
standalone Java probe first (eastward, westward, repeated crossings, and a track
hugging the edge without crossing), then as Kotlin tests: restoring the old
single-point behaviour fails four of them, and the current code is green
alongside :core:domain:test and :feature:roaming:testDebugUnitTest.
Also drops the misleading "left/right terminal position" comments: the branch
that fires when the previous sample sat near +180 is the eastward crossing, and
it correctly closes on +180.
The decompiled per-edge branches computing the surrounding nine squares had two
independent defects.
1. Field letters stepped past the alphabet
Every branch moved a field with raw character arithmetic (`str[0] - 1`,
`str5[0] + 1`) and Maidenhead fields only run A..R, so coordinates near the
edges of the world produced squares outside the alphabet:
(-89.9, -179.9) -> [@A91, AA01, AA11, @A90, AA00, @A10, @@99, A@09, A@19]
( 89.9, 179.9) -> [RS80, RS90, SS00, RR89, RR99, SR09, RR88, RR98, SR08]
2. Some moved cells kept the old field letter
The north-edge branch advanced the latitude field for the top-centre cell only,
leaving the two top corners in the previous field:
centre AA19 -> ported [AA00, AB10, AA20, ...]
correct [AB00, AB10, AB20, ...]
Cross-checked against the shared qthNeighbors helper, which is already covered
by QthConverterTest including the AA00 and RR99 wrap cases:
before: 64,800 sampled coordinates, 6,480 disagreed (all with centre square
digits 00 or x9, i.e. the north edge and the 00 corner)
after: 64,800 sampled coordinates, 0 disagree
The ring is plain Maidenhead arithmetic with no QTH-Locator-specific behaviour,
so call qthNeighbors instead of keeping a second, wrong implementation. The
now-unreferenced buildGrids branches are removed (grep confirmed the definition
was the only remaining occurrence). The existing OL42 reference grid and the
four ported edge-case tests still pass unchanged.
Reverting the fix fails both new regression tests; with it
:feature:roaming:testDebugUnitTest and :core:domain:test are green.
The QTH Locator port keeps the decompiled range tables, which close both
adjacent cells (`-20.0..0.0` then `0.0..20.0`). Kotlin's `when` takes the first
match, so any coordinate landing exactly on a field, square or subsquare
boundary was attributed to the previous cell:
(0, 0) II99xx99 should be JJ00aa00
(1, 1) JJ00lx99 should be JJ01ma00
(22, 108) OL31xx99 should be OL42aa00
(22.5, 108.5) OL42fl99 should be OL42gm00
(22.25, 108.25) OL42cf99 should be OL42dg00
At the field level the locator is wrong by a whole 20 deg x 10 deg field, and
the 3x3 neighbour grid plus the red position marker are derived from the same
characters, so the whole Roaming screen pointed at the wrong square.
Cross-checking the port against core/domain positionToQth over the grid:
before: 65,341 sampled points, 4 agreed
after: 65,341 sampled points, all agree
The independent converter was confirmed correct first: it reproduces the
user-verified reference sample OL42ih45, and hand-computing lon=-179.75
(0.25 deg into the field, x12 -> subsquare index 3 = 'd') and lat=-90
(subsquare 'a', extended digit 0) matches it rather than the port.
Rather than rewriting the faithful lookup tables, nudge the input by 1e-10 so
the closed ranges behave like the standard half-open [low, high) cells, keeping
+90/+180 inside the final R cell. Seven real-world city samples and all existing
ported-behaviour tests, including (90, 180) -> RR99xx99, are unchanged.
Regression tests added for the boundary cases and for cross-implementation
agreement. Reverting the fix fails both; with the fix
:feature:roaming:testDebugUnitTest is green.
Two races shared the same cause: pushSamples runs on the audio capture thread
while clear() runs on the Compose main thread.
1. Lost redraw notifications
Both paths did `_revision.value += 1`. That expands to get -> add -> set and is
not atomic. A controlled two-thread probe (20k increments each, five runs)
lost up to 6,402 increments / 16%; using StateFlow.update lost zero. Since
revision is the Canvas's only redraw signal, every lost update can leave the
waterfall showing stale rows. If both writes land on the same number, StateFlow
sees no value change and notifies nobody.
Use `_revision.update { it + 1 }` in both paths.
2. Clear resurrected pre-clear audio
pushSamples copies pending audio under the lock, deliberately performs FFT
outside it, then reacquires the lock to append rows. The exact interleaving:
audio thread: take old audio, start FFT
main thread: user taps Clear -> rows/pending empty
audio thread: old FFT completes -> appends old rows again
The display becomes empty then immediately redraws the audio the user cleared.
A deterministic thread probe reproduced old rows after clear. Add a generation
counter protected by the same lock: pushSamples records it before FFT and drops
the computed rows when clear incremented it meanwhile. Fixed probe remains empty.
Verification: :feature:cw:compileReleaseKotlin + full :core:domain:test BUILD
SUCCESSFUL; grep confirms no non-atomic revision increments remain.
AprsForegroundService refuses to run when callsign is blank and calls
stopSelf(), but it never writes enabled=false back to AprsStore. AprsCard did
the opposite: toggling Enable first persisted enabled=true, then started the
service. On a fresh install with no callsign this produced a permanent lie:
UI switch: ON SharedPreferences: enabled=true service: stopped
Leaving and reopening settings still showed ON even though APRS had never sent
a packet. Startup/restore code could then repeatedly try to launch a service
that immediately stops itself.
There were two entry paths with the same root cause:
1. Turning the switch on before entering a callsign.
2. Erasing an existing callsign in the dialog while APRS was already enabled.
The switch now opens the configuration dialog without persisting or starting
anything when callsign is blank. Saving the dialog also forces enabled=false
when the callsign was erased.
State-machine simulation: old state ends ON/stopped; both fixed paths end in a
consistent OFF/stopped state. :feature:settings:compileReleaseKotlin and full
:core:domain:test BUILD SUCCESSFUL.
AmSatParser deliberately uses getOrNull + mapNotNull while reading each day's
12 slots, so a shortened HTML row can legitimately produce SatDay(slots=[]).
StatusRow then selected the first non-gray slot and fell back to slots.first(),
which throws NoSuchElementException and crashes the entire AMSAT status screen.
Use firstOrNull for both lookups and render a zero-count gray placeholder when
no slot exists. Real amsat.org HTML currently has all 41 satellite rows at the
full 73 cells, but the parser's own tolerance contract means the UI must handle
what it can emit.
Verified against the live page: parser matches 41/41 rows and 477/477 reports;
:feature:status:compileReleaseKotlin BUILD SUCCESSFUL.
Without key(entry.id), Compose reuses component state by position.
When the list reorders mid-countdown (new QSO inserted at index 0,
or QRZ grid backfill triggers refreshTick++), the pending deletion
transfers to a different record and removes the wrong one.
Affected screens: LogTab and WavelogLogScreen.
mckero